One approach to get replicated results from forward-time sims

This is just a quick example of how to replicate simulations many times.

First I’m creating a ‘run_rep()’ function that takes the input parameters (see ’trad-demo-sim.Rmd`) and returns a single simulation result.

Then I use mcapply() to run this on multiple cores. This only works with linux and mac (I think). Windows will need the more elaborate parLapply() approach.

run_rep <- function(dens.scale=0.05,
                    h=225,
                    refs=c(1),
                    sizeref=c(25),
                    mix=0.01,
                    longmean=3,
                    shortscale=0.35)
{

  l <- recolonizeLandscape(dens.scale=dens.scale,
                             h=h,
                             refs=refs,
                             sizeref=sizeref,
                             mix=mix,
                             longmean=longmean,
                             shortscale=shortscale
                             )
  getpophist(l)
}

Load up some essentials

library(rmetasim)
library(parallel)
source("make-landscape.R")
source("plothist.R")
source("getpophist.R")

set.seed(123463)

run the simulation. I’m just going to run 4 reps of the same (default) parameters:

#change cache to F in the chunk header to rerun sims every time you knit
treats = expand.grid(rep=1:4,
                     dens.scale=0.05,
                    h=225,
                    refs=c(1),
                    sizeref=c(25),
                    mix=c(0.01),
                    longmean=3,
                    shortscale=0.35)

reps <- mclapply(1:dim(treats)[1],mc.cores=4,function(x)
  {
    
    with(treats,run_rep(dens.scale=dens.scale[x],
                             h=h[x],
                             refs=refs[x],
                             sizeref=sizeref[x],
                             mix=mix[x],
                             longmean=longmean[x],
                             shortscale=shortscale[x]))
  })

Here are summary plots for each rep

sapply(reps,function(x){plothist(x)})

## [1] 1 1 1 1

This one does two reps of one mixture and two of another

treats = expand.grid(rep=1:2,
                     dens.scale=0.05,
                    h=225,
                    refs=c(1),
                    sizeref=c(25),
                    mix=c(0.01,0.001),
                    longmean=3,
                    shortscale=0.35)

reps <- mclapply(1:dim(treats)[1],mc.cores=4,function(x)
  {
    
    with(treats,run_rep(dens.scale=dens.scale[x],
                             h=h[x],
                             refs=refs[x],
                             sizeref=sizeref[x],
                             mix=mix[x],
                             longmean=longmean[x],
                             shortscale=shortscale[x]))
  })

sapply(reps,function(x){plothist(x)})

## [1] 1 1 1 1

Finally, one with two different refuges, 2 reps each

treats = expand.grid(rep=1:2,
                     dens.scale=0.05,
                    h=225,
                    refs=c(1,14),
                    sizeref=c(25),
                    mix=c(0.01),
                    longmean=3,
                    shortscale=0.35)

reps <- mclapply(1:dim(treats)[1],mc.cores=4,function(x)
  {
    
    with(treats,run_rep(dens.scale=dens.scale[x],
                             h=h[x],
                             refs=refs[x],
                             sizeref=sizeref[x],
                             mix=mix[x],
                             longmean=longmean[x],
                             shortscale=shortscale[x]))
  })

sapply(reps,function(x){plothist(x)})

## [1] 1 1 1 1